Generic, Model-Based Estimation and Detection of Peaks in Image Surfaces

Trihedrals give rise to point scatterers that are important features in synthetic-aperture radar (SAR) imagery. Shadows and extended features are analyzed in terms of point scatterers. Images of point scatterers are delta function peaks that are imaged as Gaussian peaks with the SAR impulse response function. Complex features and clusters in digital images may contain peaks and valleys at diierent scales. This paper describes two new peak detection algorithms for Gaussian peaks. The statistical performances of two new peak detection operators are analyzed theoretically. Analytic results are compared with truth in synthesized images with Gaus-sian peaks and truth in SAR images. The two operators estimate accurately the peak position, peak amplitude and peak width in both of synthesized and SAR images. Experimental results obtained by applying these two peak detectors are presented.